A predictive framework in healthcare: Case study on cardiac arrest prediction

Area of Science:

• Healthcare Analytics
• Medical Informatics
• Predictive Modeling

Background:

• Data-driven healthcare leverages predictive analytics for enhanced decision-making and personalized patient care.
• Prognostic models are crucial in medical environments, with machine learning techniques widely applied.
• A lack of standardized frameworks for selecting prediction models hinders optimal application in diverse medical scenarios.

Purpose of the Study:

• To propose the ISAF framework for selecting appropriate prediction models based on classification method properties.
• To develop and validate a prognostic model for predicting cardiac arrests in sepsis patients using the ISAF framework.

Main Methods:

• Development of the ISAF (Intelligent Selection of Appropriate Framework) framework for guiding prediction model selection.
• Step-by-step application of the ISAF framework to create a prognostic model for cardiac arrest in sepsis.
• Utilized a modified stacking model for enhanced predictive performance.

Main Results:

• The developed prognostic model achieved high sensitivity in predicting cardiac arrests: 85% one hour prior (sensitivity >= 0.85) and 73% 25 hours prior (sensitivity >= 0.73).
• The model demonstrated significant improvements over the APACHE II and MEWS standard systems in prediction accuracy.
• The proposed model extended the prediction interval and enhanced performance metrics compared to prior research.

Conclusions:

• The ISAF framework provides a structured approach for selecting suitable prediction models in healthcare.
• The novel prognostic model offers a substantial advancement in early cardiac arrest detection for sepsis patients.
• This research highlights the potential for improved patient outcomes through advanced data-driven predictive analytics in critical care settings.